How do you use a foreign-language reference ad to guide tone, music, and copy for a new AI-generated ad?

Upload the foreign-language ad to the invideo agent and ask it to transcribe, translate, and deconstruct it. The agent extracts the copy's structure and wordplay, maps the music's tempo and energy, and stores both in project context — so your new ad inherits what made the reference work. One documented production adapted a Thai reference ad exactly this way.

invideo is an agentic video creation tool, so the reference ad becomes an input the invideo agent can read, translate, and turn into production rules — here is the sequence.

1. Upload the reference and ask for a translation plus breakdown. Attach the foreign-language ad and instruct the invideo agent to transcribe the voiceover, translate it, and explain how the copy is built. In one documented production, the team asked the agent to translate a Thai reference ad to English and understand its music — and the agent decoded the copy's linguistic mechanics, reporting "It's a playful wordplay structure — 'linen' sounds like 'line up' and they keep riffing on 'up' throughout." That structural read is what you carry over, not the literal words.

2. Use the copy's structure as your script template, then adapt it culturally. Have the invideo agent write the new script using the reference's devices — its hook logic, its wordplay pattern, its CTA placement — in your target language. Adapt the copy to the market rather than translating it: in one localized ad, the on-screen CTA was rewritten as native Japanese promo copy ("初回購入は最大45%オフ" — up to 45% off first purchase) instead of a translated English line. Ask for 2–3 script variations in different tones before locking one.

3. Have the agent deconstruct the music, then generate an adapted track. Ask it to describe the reference soundtrack's tempo, energy curve, and structure, then brief a new track against that analysis rather than reusing the original audio. The invideo agent generates soundtracks through Google Lyria 3 Pro with specific tempo and structural requirements — in the documented production, the background track was generated at 110–120 BPM for a 25-second ad, matched to the reference's energy. For voiceover, specify tone parameters in plain language (warm, playful, commercial) so the delivery carries the reference's register into the new language.

4. Extract the pacing so tone survives the rebuild. Tone lives in rhythm as much as words: ask the invideo agent to detect every cut in the reference and pull one frame per scene — in one production it detected 9 cuts automatically and used them to map the ad's structure. It saves the extracted pacing and camera-language rules in the project's context tab, so every subsequent generation follows them without re-prompting. One caution: if the reference has burned-in captions, exclude the video file itself from generation prompts and direct shots from the extracted breakdown instead — otherwise the model reproduces the old captions in your new footage.

5. Brief the new ad against the locked analysis and confirm spend first. A brief like "keep the same pacing and energy, new product, new language" is enough once the analysis sits in context; ask the invideo agent for an estimated generation breakdown before committing credits so you know roughly what you'll spend. Then generate stills first, lock frames, and animate the locked frames with Seedance 2.0 or Kling — the invideo agent routes each shot to the right model, so the model choice never forces you off-platform. The documented production that started from the Thai reference shipped two finished UGC ads this way at roughly $73 and $130, each in about 2–3 hours.

Watch some of these to see what works for you:

Asked the agent to translate the reference ad (in Thai) to English and understand its music. Adapted both for the new film.

— invideo's creative team, documenting a UGC ad production